Cell-Receptor Transgenic Mice Involves Deletion of Nonmature CD4+8+ Thymocytes Pawel Kisielow"T, Horst Bliithmann*, Uwe D

~74~2------ARfiCLES------~N~A~T~U~R~E~v~c~Jl~.. ~3~33~2='~·~'t~JN~E~t~9=88 Tolerance in T -cell-receptor transgenic mice involves deletion of nonmature CD4+8+ thymocytes Pawel Kisielow"t, Horst Bliithmann*, Uwe D. Staerz", Michael Steinmetz* & Harald von Boehmer" * Basel Institute for Immunology, CH-4005 Basel, Switzerland :j: Central Research Units, F. Hoffmann-La Roche & Co. Ltd, CH-4002 Basel, Switzerland

The mechanism of self-tolerance is studied in T-cell-receptor transgenic mice expressing a receptor in many of their T cells for the male (H- Y) antigen in the context of class I H-2Db MHC antigens. Autospecific T cells are deleted in male mice. The deletion affects only transgene-expressing cells with a relatively high surface-density of CDS molecules, including nonmature CD4+CD8+ thymocytes, and is not caused by anti-idiotype cells.

T LYMPHOCYTES recognize antigens on the surface of other products were absent from the pool of peripheral T cells and 12 14 cells in the context of molecules encoded by the major histocom medullary thymocytes - , but were present among cortical 1 12 13 patibility complex (MHC) by virtue of the heterodimeric T cell CD4+8+ thymocytes • . These results can be explained by receptor (TRC) which is composed of a and {3 polypeptide deletion of autospecific cells, but the alternative possibility that 2 3 chains • • In binding to its ligand, the a{3TCR is assisted by their absence is the result of a change of their phenotype caused 4 5 CD8 or CD4 accessory molecules • , which presumably interact by modulation or masking of surface molecules has not been with nonpolymorphic portions of class I or class II MHC excluded. 10 molecules respectivel/- • Mature T lymphocytes usually do The development of transgenic mice offers another approach not respond to self-MHC molecules presenting self-antigens. to analyse the mechanism of self-tolerance. To this end we have The question of whether the mechanism of immunological toler constructed transgenic mice expressing in a large fraction of ance involved deletion of autospecific lymphocytes has con their T cells an a{3TCR specific for a minor histocombatibility 1 cerned immunologists over decades \ but no direct evidence antigen (H-Y) present on male, but not female, cells. Fertilized for such a mechanism has been obtained, because the great eggs obtained from a cross of C57BL/6J x DBA/21 mice were diversity of receptors generated during lymphocyte development injected with genomic DNA harbouring the productively had made it impossible to follow individual clones of cells rearranged TCR a and {3 genes isolated from the B6.2.16 15 expressing receptors specific for self-antigens. cytolytic T-cell clone • This clone is specific for H-Y antigen Recently, two groups of investigators obtained monoclonal in the context of class I (H-2Db) MHC antigen and expresses antibodies (mAb) against the products of certain V{3 genes that a TCR {3-chain coded in part by the V{38.2 gene segment which 6 are expressed with unusually high frequency on T cells specific can be identified by the F23.1 antibodi . 12 14 for certain class II MHC-associated alloantigens - • Using The transgenic founder mouse 71 contained four copies of these antibodies, Kappler et al. and MacDonald et al. were able the a and two copies of the {3 transgenes integrated on the same 17 to show that in mice expressing the relevant class II MHC chromosome • It was crossed with C57L mice expressing H-2b associated antigens, cells expressing the particular V{3 gene MHC antigens, but lacking the V{38 gene family. Here we show that cells with the phenotype of the B6.2.16 clone that responded t Pennanent address: The Institute of Immunology and Experimental Therapy, Polish Academy to H-Y antigen were frequent in female but not in male trans of Sciences, Wroclaw, Poland. genic offspring, despite the fact that peripheral T cells in animals

Table 1 Frequency of male (H-Y) antigen-specific precursors of proliferating T cells (PT-P) among CD4+8- and CD4-8+ T cells from normal and a/3TCR transgenic mice

Donor of responding T cells af3TCR transgenic C57L Stimulation: female female male spleen cells CD4/CD8 (3000R) + IL-2 phenotype 1/frequency p* 1/frequency p !/frequency p C57BL/6 female CD4-8+ >25,000 >25,000 >25,000 C57BL/6 female CD4-8+ 2.3 (1.6-3.3) 0.66 1.8 (1.2-2.6) 0.87 2.3 (1.3-4.0) 0.18 +Con A CD4+8- NTt 6.4 (4.5-9.1) 0.60 NTt C57BL/6 male CD4-8+ 15,985 (5,029-50,802) 0.61 6.6 (4.8-9.0) 0.67 >25,000 CD4+8- NTt >25,000 NTt

Lymph node cells were stained with a mixture of anti CD4-PE and anti-CD8-FITC mabs (see Fig. 1). CD4+8- and CD4-CD8+ T cells were separated on fluorescein activated cell sorter (FACS440, Becton Dickinson). Limiting numbers of CD4+8- CD4-8+ T cells (24 wells per group) were cultured for 8 days together with irradiated (3,000R) spleen cells (5 x 105 cells per cell) and interleukin-w (5% v/v) of partially purified 24 1 supernatant from Con A-stimulated rat spleen cells without or with Con A (2.5 f.Lg ml- ). Cells were collected after addition of eHJthymidine for the last 12 h of culture and incorporated radioactivity was measured by liquid scintillation counting. Negative control cultures contained no 25 responder cells. Frequencies were calculated as described elsewhere • * Probability, p, attached to the computed x2 (ref. 25). t NT, not tested.

17 of both sexes expressed both transgenes • T cells in male (but Littermates Lymph node cells not female) mice had an abnormal CD4/CD8 phenotype: over 90% of T cells in male transgenic mice were CD4-s-, or a expressed only low levels of CDS molecules, and the numbers of CD4+8-T cells were very small. The cellular composition of non transgenic male thymuses revealed that this unusual phenotype of male peripheral T cells was the consequence of deletion of auto specific thymocytes expressing high levels of CDS molecules, predominantly cortical CD4+8+ thymocytes. The deletion pro Q) c0 b cess spared cells expressing low levels of CDS molecules, but Q) UJ UJ 0 a.. a.. affected the precursors of single positive CD4+s- cells that were (/) transgenic !!' 0 T cells in females ....J c Lymph nodes of female transgenic mice contained normal pro portions of CD4+8- and CD4-s+ T (Thy!+) cells which had transgenic normal levels of CD4 as well as CDS accessory molecules (Fig. male 1 a, b, d and e). But these differed in two respects from T cells in normal mice. Firstly, as previously described for~ transgenic 15 mice , most of them expressed the transgenic~ chain on their anti-Thy 1 ( FITC) anti-CD8 ( FITC) surface (Fig. 1 b). Secondly, as shown by limiting dilution analy Log green fluorescence sis of CD4-8+ T cells, one in six proliferated specifically in response to C57BL/ 6 male stimulator cells, as compared with Fig. 1 Comparison of cell surface expression of F23.1 + TCR f3 one in 16,000 in normal C57L female mice (see Table 1). As chain, CD4 and CDS molecules on lymph node T (Thy!+) cells only every second plated T cell responded to concanavalin A from female and male af3TCR transgenic mice and their nontrans (Con A), we conclude that at least 30% of CD4-s+ T cells in genic male littermate as analysed by two-colour flow cytometry. Lymph node cells were stained with biotinylated F23.1 monoclonal transgenic females have a phenotype similar to that of the antibody (mAb) followed by a mixture of fluorescein (FITC) B6.2.16 clone. CD4+8- T cells from transgenic mice did not labelled anti-Thy! mab with phycoerythrin-streptavidin (PEA) (a, show any male-specific proliferation, but did respond to Con band c) or with PE-conjugated anti-CD4 mAb followed by FITC A (Table 1). conjugated anti-CD8 mab ( d, e and fl. In panels a, b and c some Thy!- cells (B cells) stain nonspecifically with F23.1 mAb due to T cells in males the binding by Fe receptor. The presented data were obtained with As in transgenic females, lymph nodes of transgenic males one pair of 7-week-old af3TCR transgenic female and male litter contained normal proportions of Thyl + cells, and most of them mates. The same results were obtained with 3 other pairs of trans genic mice. expressed the transgenic ~ chain on their surface (Fig. lc). Methods. Single cell suspensions were prepared from lymph nodes Northern blot analysis of the a trans gene revealed comparable 17 (mesenteric, axillary, inguinal) and washed twice in RPM1-1640 levels of expression in T cells from female and male mice • and once in PBS with 5% FCS. For staining the following mAbs However, the CD4/CD8 phenotype ofT cells in male mice was were used: FITC-conjugated anti-Thy! (ref. 21), biotin-conjugated very different from that of females: 58% of Thy!+ cells were F23.1 (ref. 16), PE-conjugated anti-CD4 (anti-mouse L3T4, Becton CD4-s-, 35% were CD4-8+ but expressed low levels of CDS, Dickinson) FITC conjugated anti-CD8 (anti-mouse Lyt2, Becton and 7% were CD4+8- and expressed normal amounts of CD4 Dickinson). Biotin or FITC conjugation of mAbs was performed (Fig. 1/). Limiting dilution analysis showed that one in two by standard procedures. Optimal concentrations of staining reagents were determined in preliminary experiments. All incuba CD4-g+ T cells could be induced to grow by Con A. There was, 6 tions and washings were done at 4 °C. Cells (0.5-1 x 10 ) were however, no detectable response to male C57BL/ 6 stimulator incubated with either biotinylated F23.1 mAb (a, b and c) or cells (Table 1). Likewise, CD4_8_ and CD4+s- T cells were anti-CD4-PE mAb (d, e and fl. After 20 min, cells were washed unresponsive to H-Y antigen (data not shown). These results twice and incubated again for 20 min with anti-Thyl-FITC mAb indicate that male-specific T cells with the phenotype of the plus PEA (Becton Dickinson) (a, band c) or with anti-CD8-FITC B6.2.16 clone are absent from male transgenic mice and that mAb (d, e and fl. Finally, cells were washed three times in PBS the lack of or low level of CDS on transgene-expressing cells 5% FCS and analysed for two-colour fluorescence on FACScan precluded male-specific responses. This conclusion is supported (Becton Dickinson) flow cytometer with a single Argon laser and by the observation that cytolytic activity of the B6.2.16 clone logarithmic intensity scales using F ACScan research software pro can easily be inhibited by anti-CD8 antibodies (not shown), gram (FRSP). Ten thousand viable cells were analysed in each sample. Dead cells were excluded from analysis using a combina and by CDS gene transfection experiments which show that 4 5 tion of low-angle and sideways light scatter. The results are presen CDS molecules strongly assist antigen recognition by T cells • • 17 ted as 'density' plots, generated by analysis of processed data As shown elsewhere , a high proportion of CD4_8_T cells in reduced to a 64 x 64 matrix with 16 levels. Percentages of stained male mice expressed both trans genes, but had their endogenous and non-stained cells were calculated using FRSP. Markers were a and ~ genes in germline configuration. This result, and the set against the 'density' plots of control samples which involved fact that only a few T cells expressed normal amounts of CDS substitution of diluent alone for either one or both antibodies. in male transgenic mice, is consistent with the notion that most T cells in male mice and CDS+ T cells in female mice carry As shown in Fig. 2a, double-staining with F23.1 and CD3 transgenic a~TCR on their surface. But, owing to reduced antibodies demonstrated that most (>95%) thymocytes from density of CDS molecules in male mice, they are not transgenic females and males expressed the ~ transgene, and autoreactive. that the amount of TCR expression corresponded to the higher values of the normal spectrum of TCR densities observed in Thymocytes C57BL/ 6 mice. The number of thymocytes was drastically lower in male (0.5- In the thymus of transgenic females, two populations express 1.6 x 107 per thymus) than in female (1.0-1.6 x 108 per thymus) ing different levels of TCR could be distinguished (Fig. 2a, transgenic mice. middle panel). The one with relatively low TCR density included

Fig. 2 Expression of F23.1+ TCR {3 Thymocytes chain, CD3, CD4, and CD8 molecules on thymocyte subpopula a b c .-'------·-----,.,-- ·-.. ------·-Tr------~ tions from normal C57BL/6 and 82.2% from a{3TCR transgenic female and male mice. In panel a, cells were incubated consecutively with anti C57BLI6 CD3 mAb, FITC-conjugated goat male anti hamster immunoglobulin, mouse immunoglobulin, biotiny lated F23.1 mAb and PEA. In panels b and c, cells were stained bith bio tinylated F23.1 mAb, followed by a mixture of anti-CD4-FITC (b) or transgenic anti-CD8-FITC (c) mAbs with PEA. female In panel d, cells were stained with anti-CD4-PE followed by anti-CD8- FITC mabs. The number of cells per thymus in C57BL/6 male, a{3TCR transgenic female and a{3TCR trans 6 6 genic male were: 100 x 10 , 105 x 10 transgenic 6 and 13 x 10 respectively. Thick male arrows indicate the population of CD4+8+ thymocytes expressing a lower level of TCR, which is mostly depleted in male thymus. Open-head anti·CD3 (FJTC) anti-CD4 (FITC) anti-CD8(FITC) anti-CD8 (FITC) arrows indicate the population of L fl res en e CD4+8-,andthinarrowsofCD4-s+ og green uo c c female thymocytes that express higher levels of TCR. Populations indicated by open-head thin arrows and in the upper left quadrant of middle panels b and c also contain CD4-CD8- thymocytes, as indicated by virtual absence of cells in the lower left quadrant of middle panel c. Presence of cells in lower left quadrant of middle panel b is due to imperfect staining of this particular sample in this experiment. In other experiments, no F23.1-CD4- cells could be seen under the same conditions. Methods. Single thymocyte suspensions were prepared by squeezing the whole thymus through a nylon mesh into medium RPM1-1640 with 5% FCS. After washing, cells were resuspended in PBS with 5% FCS, counted and stained as indicated above with extensive washings between each step (see Fig. 1). For staining with anti-CD3, unconjugated mAb 145.2c11 (ref. 22) was used. To saturate free binding-sites of second-step 1 reagent, cells were incubated with mouse immunoglobulin (Sigma, 1 mg ml- ) for 15 min. FITC-conjugated anti-CD4 (GK1.5, ref. 23) mAb was prepared by standard procedures. Control samples were stained with each reagent alone, or in combinations omitting each single reagent. Ten thousand viable cells were analysed in each sample by FACScan flow cytometry. For details see Fig. 1.

CD4+s+ cells, whereas the other, expressing about tenfold more bone marrow cells (BMC) from transgenic females were trans TCR, contained CD4+s-, CD4-s+ and CD4·s- cells (Fig. 2b ferred either along or together with BMC from normal C57L and c, middle panels). females into lethally X-irradiated female and male C57L Double-staining with CD4 and CDS antibodies revealed sig recipients. Five weeks after the transfer of the transgenic BMC, nificant differences between transgenic and normal C57BL/6 the cellular composition of the thymus in male recipients was females with regard to the size ofCD4·s+, CD4+8+ and CD4_8_ very much like that in male transgenic mice (Fig. 3a, lower thymocyte subpopulations (Fig. 2d, upper and middle panels). panel). But in the thymus of male recipients which had received The proportion of CD4+s- thymocytes in transgenic females a mixture of BMC from transgenic and C57L females, CD4+s+ was normal, but the proportion of CD4-8+ thymocytes was thymocytes derived from F23.1- C57L donors developed nor enlarged, resulting in a reversed ratio of CD4+8- to CD4·s+ mally and outgrew the transgenic F23.1 + cells, which were cells as compared with normal nontransgenic mice. The propor mostly deleted (Fig. 3b and c, lower panel). On the other hand, tion of CD4_8_ thymocytes was also noticeably higher in trans in the female recipient, CD4+8+ thymocytes developed from genic females than in normal mice. The increase in proportion both F23.1+ and F23.1- donors (Fig. 3b and c, upper panel). of CD4·s+ and CD4_8_ cells was matched by a corresponding Thus, because the deletion selectively affected transgene decrease in the size of the CD4+8+ population. expressing F23.t+ CD4+s+ thymocytes in the male recipients, In contrast to the females, the thymus of transgenic males this experiment indicates that the deletion is a result of the was severely depleted ofCD4+8+ cells with decreased expression interaction of autospecific thymocytes with radioresistant male TCR, but contained about the same total number of CD4_8_ cells in the thymus, and not of stress and steroid release. If the cells, which constituted the bulk of the population of male latter possibility were true, the F23.1- CD4+s+ thymocytes thymocytes (Fig. 2d, middle and lower panels). Most CD4·s+ derived from C57L donors of BMC should also have been cells showed low expression of CDS. Thus, the male thymus affected. was depleted of transgene-expressing cells with relatively high levels of CD4/ CDS accessory molecules and the nonmature Discussion CD4+8+ thymocytes expressing decreased amounts ofTCR were Our study with aJ3TCR transgenic mice provides new observa the main target of depletion. tions relevant to the understanding of the mechanism of self Because CD4+s+ thymocytes are extremely steroid-sensitive, tolerance and relevant to the clarification of the function of it was important to find out whether the deletion of these cells cortical CD4+s+ thymocytes in T-cell development. The drasti in transgenic males was a result of stress rather than of an cally decreased number of thymocytes in male but not female antigen-specific deletion process. Stress in the male mice could mice is direct evidence for a deletion of autospecific cells at the possibly be caused by autoimmunity not detectable by in vitro level of CD4+s+ nonmature thymocytes. Furthermore, our assay. We addressed this question in reconstitution experiments results indicate that CD4 and CDS accessory molecules are using haemopoietic stem cells from transgenic (F23.1 +) and involved in the deletion process of autospecific cells. nontransgenic C57L (F23.1-) mice (Fig. 3 ). T-cell-depleted Two questions relating to the function of double-positive

Fig. 3 Surface phenotype of thymocytes from Recipient (850R) Thymocytes C57L male and female irradiation chimeras of bone marrow cells• reconstituted with bone marrow of a,BTCR trans b c genic female, either alone (a) or with normal 50.0% bone marrow from C57L female (b, c). Thy mocytes were stained with anti-CD4-PE (a, b) or biotinylated F23.1 (c) mAbs, followed by anti CD8-FITC mAb (a, b) or a mixture of CD8- C57L female FITC mAb with PEA (c). In (a), 80% of thy UJ mocytes in the female recipient and 67% in the a. male recipient were stained with F23 .1 mAb (data not shown). 738% Methods. Bone-marrow cells from transgenic or normal C57L dnor were treated with cytotoxic anti Thyl mAb (T24, ref. 21) plus rabbit comple C57L male ment (Cedar Lane, Ontario, Canada) for 45 min at 37 °C. After washing, 5 x 106 viable cells from the transgenic donor were injected intravenously (i.v.) into lethally irradiated (850R) eight-week old C57L females and males, either alone or together with 0.5 x 106 viable bone marrow cells anti-CD8(FITC) from normal female C57L. Five weeks later the Log green fluorescence mice were killed, their thymuses removed and single-cell suspensions prepared, counted and stained with anti-CD4, -CDS and -F23.1 mAb, and analysed as described in Figs 1 and 2.

CD4+s+ thymocytes are why so many of these cells should die would be difficult to detect. Another possible reason for the within the thymus18 and whether or not they contain precursors different findings is the fact that in our transgenic mice the 19 of single positive CD4+s- and CD4-s+ cells . Our results show expression of TCR proteins is skewed towards higher levels of that the death of at least some cortical thymocytes can result the range observed in CD4+s+ from normal mice. This from antigen-specific elimination of autoreactive cells. The dele phenomenon, as well as the increased proportion of CD4-s+ tion of nonfunctional, antigen-specific CD4+s+ thymocytes thymocytes in transgenic females, could reflect a positive selec would make sense if CD4+s+ thymocytes contained precursors tion of thymocytes by H-2b antigens, or alternatively may be a of functional CD4+s- and CD4-s+ cells. Consistent with this direct consequence of expression of transgenes. We could argue view is our observation that CD4+s- cells were severely depleted therefore that the deletion of CD4+s+ thymocytes was easily in male transgenic mice, despite the fact that such cells from detected because the majority of CD4+s+ thymocytes in trans transgenic female mice cannot be activated by male cells. An genic mice represent a minor and more mature population of 14 analogous finding has been reported by MacDonald et a/. , CD4+s+ thymocytes that may escape detection in normal mice, who observed that CD4+s- and CD4-s+ T cells staining with especially when representing only a fraction of cells expressing V13 6 antibodies were reduced to the same extent in animals a certain idiotype. Whatever the reason for the apparent dis positive for the MJs•-allele of the minor lymphocyte stimulating crepancy in the results, our data indicate that the nonmature locus (Mls), even though CD4-s+ from MJs•-negative animals CD4+s+ populaton can be a target of deletion, whereas it is not lacked specificity for MJs•. clear whether CD4+s- or CD4-s+ cells are susceptible to the We thus favour the view that at least some double-positive same deletion process. CD4+s+ thymocytes act as precursors for functional single posi The second difference between our results and those of 20 12 13 14 tive cells , even though further investigation is needed. Kappler et a/. • and MacDonald et a/. is that these authors Although we have shown that the deletion predominantly affects did not report the presence of cells with few or no accessory CD4+s+ thymocytes, we could argue that it might occur indepen molecules, spared by the deletion. Again in this case, such cells dently of accessory molecules at any stage ofT-cell development. would constitute a very minor population in their experimental But this view is not compatible with our observation that the system, because the pool ofT cells in normal mice can be easily deletion process spares T cells that lack accessory molecules, replenished by T cells expressing different TRCs, which is not or even T cells having low expression of CDS. Thus our experi the case in transgenic mice. ments provide the first direct evidence that these molecules play As we observed normal numbers ofT cells in the periphery, a crucial role in the induction of tolerance. Taken together, the but not in the thymus, of transgenic males, we propose that the three observations made in male transgenic mice, namely the number of peripheral T cells can be adjusted independently of drastically reduced number of CD4+s+ thymocytes, the reduc the export of newly formed cells from the thymus. This would tion ofCD4+s- cells and the occurrence oftransgene-expressing allow the accumulation of cells with rare phenotypes in the cells with virtually no CDS, argue that the deletion of auto periphery of male mice, as shown here and in the accompanying 17 specific cells is dependent on CD4 and CDS accessory paper . The fact that transgene-expressing cells with few or no molecules. accessory molecules accumulate in male mice, tends to rule out The results of our experiments with transgenic mice differ in a role of anti-idiotypic cells in the deletion process; such a 12 14 at least two important aspects from others recently reported - mechanism should eliminate transgene-expressing cells, rather 12 13 for normal mice. Firstly, in the experiments of Kappler et a/. • , than cells expressing high levels of accessory molecules. the depletion of autospecific T cells did not appear to affect We would like to thank Katrin Hafen, Abigail Garner and CD4+s+ thymocytes. One possible reason for the difference is Verena Stauffer for technical assistance, Drs Werner Haas and that different antigens are under investigation: we are looking Hung Sia Teh for critical reading and S. Kirschbaum for typing at an antigen in the context of class I MHC antigens found the manuscript. The Basel Institute for Immunology was foun 12 13 throughout the cortex whereas Kappler et a/. • are looking at ded and is supported by F. Hoffman-LaRoche & Co. Ltd, Basel, an entity21 related to class II MHC antigens which are usually Switzerland. not detected in the outer cortex. Thus in the latter case CD4+s+ cells can meet antigen only when reaching the cortico-medullary Received 7 April; accepted 6 May 1988. junction. Consequently only a minor subset of CD4+s+ cells I. Schwartz, R. M. A. Rev. lmmun. 3, 237-261 (1985). would be deleted in the experiments of Kappler et al., and this 2. Dembic, Z. et al. Nature 320, 232-238 {1986).

3. Saito, T., Weiss, A., Miller, J., Norcross, M.A. & Germain, R.N. Nature 32S, 125-130 15 Uematsu, Y et al. Cell (in the press). (1987). 16 Staerz, U., Rammensee, H., Benedetto, J. & Bevan, M. J. Jmmun. 134, 3994-4000 (1985). 4. Dembic, Z. et aL Nature 326, 510-511 (1987). 17. Bliithmann, M. et al. Nature, submitted. 5. Gabert, J. et a/. Cell 50, 545-554 (1987). 18. McPhee, D., Pye, J. & Shortman, K. Thymus I, 151-162 (1979). 6. Swain, S. L. Proc. natn. Acad. Sci. U.S.A. 78,7101-7105 (1981). 19. Shortmann, K. & Scollay, R. in Recognition and Regulation in Ceff.mediated Immunity (eds 7. Marrack, P. eta/. J. exp. Med. 158, 1077-1091 (1983). Watson, J.D. & Marbrook, J.) 31-60 (Dekker; New York and Basel, 1985). 8. Biddison, W. E., Rao, P. E., Talle, M. A., Goldstein, G. & Shaw, S. J. exp. Med. 156, 20. von Boehmer, H. ImmunoL Today Vol? 333-336 (1986). 1065-1076 (1982). 21. Dennert, G., Hyman, R., Lesley, J. & Trowbridge, I. S. Cell Immun. 53,350-364 (1980). 9. Gay, D. eta/. Nature 328,626-629 (1987). 22. Leo, 0., Foo, M., Sachs, D. M., Samelson, L. E. & Bluestone, J. A. Proc. natn. Acad. Sci. 10. Doyle, C. & Strominger, J. L. Nature 330, 256-259 (1987). U.S.A. R4, 1374-1378 (1987). 11. Burnet, F. M. The Clonal Selection Theory of Acquired Immunity (Cambridge University 23. Dialynas, D. P. et al. lmmunol. Rev. 74, 29-56 (1983). Press, 1959). 24. Schreier. M. H., Iscove, N. N., Tees, R., Aarden, L. & von Boehmer, H. lmmunol. Rev. 51, 12. Kappler, J. W., Roehm, N. & Marrack, P. Cell49, 273-280 (1987). 115-33fi (1980) 13. Kappler, J. W., Staerz, U. D., White, J. & Marrack, P. C. Nature 332, 35-40 (1988). 25. Fazekas de St. Groth, S. J. 1mmunol. Meth. 49, Rll-R23 (1982). 14. MacDonald, H. R eta{ Nature 332, 40-45 (1988). 26. Marrack. P. & Kappler, J. Nature 332,840 (1988).

~--~~~~--LETIERSTO NATURE--

Spin-down of the X-ray pulsar GXl +4 600.------, during an extended low state Ul :::!: 400 cUl K. Makishima*, T. Ohashi*, T. Sakao*, T. Dotanit, ~ 0 u 200 ~~~~!V1M~ H. Inouet, K. Koyamat, F. Makinot, K. Mitsudat, p;~lci<9r.ol.ir1Ci ·--·· ·· ·-·-· ---· --···- -·--· ·-- ·- -· -· ·- · F. Nagaset, H. D. Thomas:j:, M. J. L. Turner:j:, T. Kii§ i & Y. Tawara§ OW---~--~~--~--~~~~~--~~ 23:30 23:40 *Department of Physics, University of Tokyo, 7-3-1 Hongo, Time ( uT, March 30, 1987) Bunkyo-ku, Tokyo 113, Japan t Institute of Space and Astronautical Science, 3-1-1 Y oshinodai, Sagamihara, Kanagawa, Japan Fig. 1 A part of the 1-20 keY count rate record during the present :j: X-ray Astronomy Group, Department of Physics, GXI +4 observation, binned at 4.0 s. The arrows indicate the University of Leicester, Leicester LEI 7RH, UK expected dip recurrence at 110.2 s. The background level, deter §Department of Astrophysics, Nagoya University, Furo-cho, Chikusa mined using the standard Ginga background estimation procedure (M.J.L.T. eta/. 1988, manuscript in preparation), should be accur- ku, Nagoya, Japan 1 ate to -2cs- •

1 2 X-ray pulsars • are magnetized, spinning neutron stars accreting matter from their binary companions. Their pulse periods P, 12 ranging over four orders of magnitude, increase and decrease in board Ginga launched on 5 February 1987 . The LAC experi 1 3 4 complex ways • • • The more luminous ones tend to show faster ment, a UK-Japan collaboration, uses eight identical propor spin-up1·s. A puzzle is that the spin-up timescales of many X-ray tional counters, achieving 4,000 cm 2 total effective area. The pulsars are much shorter than their binary-evolution timescales, LAC was pointed to GXI +4 from 07:24 to 09:08 UT on 28 thus apparently violating the steady-state condition. It has there March, and I6:52 on 30 March to 03:02 on 3I March I987. The fore been suspected6 that there exist many 'turned-off' X-ray available data consisted of IO separate segments of -20 min pulsars currently spinning down undetected. An excellent test for each. The data covered an energy range of I-36 keY with 48 this hypothesis became available using the X-ray pulsar GXl +4, spectral channels and 0.5-s time resolution. Figure I shows a 1 7 10 which used to show the fastest spin-up over a decade • - and then portion of the raw X-ray count-rate record; the source was 11 12 1 faded away • Using the X-ray satellite Ginga , we detected unambiguously detected at about 33 c s- (1-30 keY), or GXl +4 at -1/40 the previous intensity, and found that it now 3 mCrab, and showed significant flickering. The pulse-phase has an average spin-down trend. This discovery apparently supports averaged X-ray spectrum is slightly softer than those observed 2 8 18 the above hypothesis. previously • • • At a IO kpc distance, the observed 2-20 keY 36 1 GXI +4 (4UI728- 24) is a luminous binary X-ray pulsar near luminosity becomes 1.5 x I0 ergs s- • 13 the Galactic centre , with a -2 min (or possibly twice this) A standard Fourier analysis revealed a highly significant 7 13 pulse period and probably a> IOO-day binary period • • From period at -II 0 s, together with several higher harmonics, but its discovery in the early I970s until I980, it was spinning up no other periodicity. Through a folding analysis, we refined the 8 1 1 rapidly, with P=-7.5xi0- ss- (P/P=-0.02yr- ) (refs 7- heliocentric period to P = 110.233 ± 0.003 s (ref. I9) with a 1 IO), the fastest among binary X-ray pulsars . The optical com reduced chi-squared of I5.4 for 99 degrees of freedom. As plotted 13 panion is an M6III symbiotic star ; it is thus a rather rare in Fig. 2, this value for Pis longer than the value I09.668 ± 0.002 s 8 system composed of an evolved late-type primary star and a last obtained in I980 • Therefore the average behaviourofGXI + strongly magnetized neutron star. Although in the I970s GX I+ 4 4 between I980 and I987 proves to be spin-down at a rate 37 38 1 9 1 had an X-ray intensity of 50-200 mCrab (10 - erg s- ; ref. 7), P = 2.6 x I0- s s- , in contrast to the previous rapid spin-up. A in I983 it was found by Exosat to be <4 mCrab (ref. 11 ). This joint Australia-Germany-Italy balloon experiment in Novem and subsequent Exosat observations (K. Mukai, personal com ber I986 obtained a similar value for P (A. B. Giles, J. Greenhill munication) showed that GXI +4 had entered an extended low and D. P. Sharma, personal communication). state ( <0.5 mCrab ), although the hard X-ray flux 14 might have The folded pulse profile (Fig. 3) turned out to be extremely remained relatively unchanged15 and the optical emission peculiar: the pulse peak involves a sharp dip, which is resolved 16 17 showed a complex time variation • • to two roughly symmetric minima separated by a local narrow The present observation was carried out with the large area maximum. On closer inspection, we can identify individual dips counter (LAC; M.J.L.T. et a/., manuscript in preparation) on in the raw data of Fig. I. Such a peculiar feature has never been